CN104077309A - Method and device for carrying out dimension reduction processing on time-sequential sequence - Google Patents
Method and device for carrying out dimension reduction processing on time-sequential sequence Download PDFInfo
- Publication number
- CN104077309A CN104077309A CN201310105366.0A CN201310105366A CN104077309A CN 104077309 A CN104077309 A CN 104077309A CN 201310105366 A CN201310105366 A CN 201310105366A CN 104077309 A CN104077309 A CN 104077309A
- Authority
- CN
- China
- Prior art keywords
- time
- time series
- distance
- duration
- time point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/283—Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a device for carrying out dimension reduction processing on a time-sequential sequence, and belongs to the technical field of a computer. The method comprises the following steps that: a time-sequential sequence to be processed is obtained; and the time-sequential sequence is subjected to PLA (Piecewise Linear Approximation) processing, and the duration of the time slice of the PLA processing is not fixed and is an integral multiple of the preset unit duration. After the method and the device are adopted, the storage space occupied for storing the time-sequential sequence can be reduced.
Description
Technical field
The present invention relates to field of computer technology, particularly a kind of method and apparatus that time series is carried out to dimension-reduction treatment.
Background technology
Along with the develop rapidly of database technology, people start to pay close attention to how from large-scale data, to obtain valuable information, and this process can be called large data analysis.In actual applications, a lot of situations are all to analyze for time series data in large data analysis.Time series data refers to time series data, is the data rows recording in chronological order under unified index, for example, and the transaction data of stock market, the status data that sensor network is collected, the consumption statistics in shop, telephony traffic statistics etc.
The data volume of time series data is very huge, for storage and the retrieval of order sequenced data at one's leisure, can take dimension-reduction treatment to time series data, the data that the data compression that is about to more time point is less time point.PLA(Piecewise Linear Approximation, piecewise linear approximation) be a kind of conventional dimension-reduction treatment method.PLA is cut into little time slice by time series data, in each time slice, with a line segment with certain slope, be similar to the data of this time slice, like this, during time series after stores processor, only need to store time point and the corresponding linear dimensions (coefficient of straight-line equation under line segment) of the initial sum termination of the line segment that each time slice is corresponding, can effectively save storage space.
Time Series Similar retrieval is a kind of analysis means conventional in large data analysis.Its way is, huge time series data is divided into the time series that a large amount of durations are equal to be stored, according to the Goal time order sequence (Goal time order sequence is identical with each time series duration of storage) of retrieval, the time series that inquiry matches with it in each time series of storage.For example, in cardiogram, the frequency of occurrences of certain signature waveform can, for judging certain disease, can be retrieved this signature waveform, and carry out diseases analysis according to result for retrieval in the cardiogram of record.For the ease of retrieval, generally time series and the Goal time order sequence of storage are all carried out to fixed length PLA processing.Fixed length PLA, in the process of processing, is cut into by time series the time slice that a plurality of durations are equal at PLA.
In realizing process of the present invention, inventor finds that prior art at least exists following problem:
In prior art, when time series is stored, carry out fixed length PLA processing, for fixed length PLA, need to by shortening the duration of time slice, guarantee the precision of data, this will increase the data volume that needs storage, larger to taking of storage space.
Summary of the invention
In order to solve the problem of prior art, the embodiment of the present invention provides a kind of method and apparatus that time series is carried out to dimension-reduction treatment, to reduce storage shared storage space during time series.Described technical scheme is as follows:
On the one hand, provide a kind of method of time series being carried out to dimension-reduction treatment, described method comprises:
Obtain pending time series;
Described time series is carried out to piecewise linear approximation PLA processing, and the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration.
On the other hand, provide a kind of method that time series is retrieved, pre-stored have to adopt described above time series is carried out to the time series that the method for dimension-reduction treatment is processed, described method comprises:
The inquiry request of Goal time order sequence is carried in reception;
The same way that employing is carried out dimension-reduction treatment to the time series of storage is carried out dimension-reduction treatment to described Goal time order sequence;
In the time series of storage, the time series that the Goal time order sequence after inquiry and processing matches.
On the other hand, provide a kind of device that time series is carried out to dimension-reduction treatment, described device comprises:
Acquisition module, for obtaining pending time series;
Processing module, for described time series being carried out to piecewise linear approximation PLA processing, the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration.
On the other hand, provide a kind of device that time series is retrieved, described device comprises:
Memory module, for the pre-stored employing time series that time series is carried out to the device processing of dimension-reduction treatment described above;
Receiver module, for receiving the inquiry request of carrying Goal time order sequence;
Processing module, for adopting the same way that the time series of storage is carried out to dimension-reduction treatment to carry out dimension-reduction treatment to described Goal time order sequence;
Enquiry module, for the time series in storage, the time series that the Goal time order sequence after inquiry and processing matches.
The beneficial effect that the technical scheme that the embodiment of the present invention provides is brought is:
In the embodiment of the present invention, adopt the duration of time slice fixing and be the PLA processing mode of the integral multiple of default unit duration, time series is carried out to dimension-reduction treatment, like this, with respect to fixed length PLA, can substitute a plurality of time slices in fixed length PLA with a time slice, thereby, storage space shared while storing time series can be reduced.
Accompanying drawing explanation
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing of required use during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.
Fig. 1 is the method flow diagram that time series is carried out to dimension-reduction treatment that the embodiment of the present invention provides;
Fig. 2 is the method flow diagram that time series is retrieved that the embodiment of the present invention provides;
Fig. 3 is the apparatus structure schematic diagram that time series is carried out to dimension-reduction treatment that the embodiment of the present invention provides;
Fig. 4 is the apparatus structure schematic diagram that time series is retrieved that the embodiment of the present invention provides.
Embodiment
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
The embodiment of the present invention provides a kind of method of time series being carried out to dimension-reduction treatment, and the method is time series to be carried out to the method for dimension-reduction treatment storage, in the database that can be applied to store a large amount of time seriess.As shown in Figure 1, the treatment scheme of the method can comprise following step:
Step 101, obtains pending time series.
With database, the time series data of real-time generation is stored as to example, As time goes on, can constantly produce new time series data, the duration that every process is certain (duration of the time series setting in advance), can obtain the time series data producing in this section of duration, as pending time series, and then can carry out the operation of follow-up dimension-reduction treatment.In addition, in this step, also can obtain the time series of having stored in database, carry out follow-up dimension-reduction treatment, or, in this step, also can obtain the Goal time order sequence of carrying out time series retrieval, carry out follow-up dimension-reduction treatment.
Step 102, carries out PLA processing to the time series obtaining, and the duration of the time slice that this PLA processes duration fixing and time slice is the integral multiple of default unit duration.
Wherein, when carrying out the arranging of unit duration, N/mono-(N can be any positive integer, and the value of N can arrange according to the requirement of degree of accuracy and processing speed) of its duration that is time series can be set.The duration of each time slice can be the arbitrary integer times (being no more than N) of unit duration, the duration of each time slice can be different, like this, in the process of processing at PLA, the duration of time slice selects degree of freedom larger, this makes this PLA process the principle of carrying out time slice division according to data variation trend that can follow as much as possible PLA, as far as possible data over time trend change and carry out time slice division (this principle at the beginning of being the design of PLA near the time point of (as become and reduce trend from increase tendency), fixed length PLA reality has been destroyed this principle to a certain extent).
After step 102, can also comprise the step that the time series after processing is stored.
Wherein, the time series after processing is the time series of the dimensionality reduction of processing through above-mentioned PLA.The content of the time series after this processing can comprise start time point and the linear dimensions of tactic each time slice chronologically.Time slice end time point is the start time point of a time slice thereafter, thus not writing time fragment end time point.The linear dimensions of time slice, can be the coefficient in the functional expression of straight line under line segment that time slice is corresponding, i.e. a and b in f (x)=ax+b, and wherein x is time variable, f (x) is the data that time point is corresponding.
In the embodiment of the present invention, for the processing procedure of above-mentioned steps 102, can there is multiple different concrete manner of execution, provide several manners of execution wherein below.
Method one, can comprise following treatment step:
Step 1, the duration that the time series obtaining is carried out to time slice is that the PLA of default unit duration processes.Wherein, the duration of time series is the integral multiple of described unit duration.
The duration that first time series is carried out to time slice in the method is that the fixed length PLA of unit duration processes, then the merging of carrying out time slice is processed.
Step 2, in each time slice of duration Wei Gai unit duration, if meet default approximate condition between line segment corresponding to a plurality of time slices that are linked in sequence, a plurality of time slices that this are linked in sequence merge processing.
Wherein, each time slice of duration Wei Gai unit duration is each time slice marking off in the fixed length PLA processing of above-mentioned steps.Approximate condition i.e. the requirement to the degree of approximation between each line segment, if meet the requirement of certain degree of approximation between line segment corresponding to a plurality of time slices that are linked in sequence, can merge these time slices.
This merges processes and can comprise: a plurality of time slices that are linked in sequence are merged into a time slice, the broken line that line segment corresponding to time slice that merging obtains forms for line segment corresponding to these a plurality of time slices of approximate representation.The time point that merges two end points of the time slice obtaining can be respectively start time point and the end time point of this time slice.The data value that merges two end points of the time slice obtaining can be determined at the data value of this start time point and this end time point according to this time series respectively, for example, can choose with the difference of the data value of this start time point and be less than the data value of certain threshold value as the data value of the first end points, choose data value that the difference with the data value of this end time point is less than certain threshold value as double-pointed data value.Preferably, the data value that merges two end points of the time slice obtaining can be respectively this time series at the data value of this start time point and this end time point, also the line that, merges outermost two end points in each end points of the line segment that a plurality of time slices that line segment corresponding to time slice obtain be linked in sequence are for this reason corresponding.
The process of processing by merging, can be reduced to one group of data corresponding to a time slice after merging by multi-group data (every group of data comprise start time point and linear dimensions) corresponding to a plurality of time slices before merging.
Concrete, the implementation of step 2 can be: at duration, be in each time slice of unit duration, if the absolute value of line segment corresponding to a plurality of time slices that are linked in sequence slope differences is each other less than default first threshold, the described a plurality of time slices that are linked in sequence are merged to processing.A plurality of time slices that are linked in sequence can be that two time slices can be also plural time slices.Here, the condition of merging can be that the absolute value of line segment that a plurality of time slices of being linked in sequence are corresponding slope differences between any two is all less than first threshold.Further, this step, in implementation process, can adopt following multiple executive mode.
Processing mode one, its concrete implementation can comprise the steps:
Steps A, is in each time slice of unit duration at duration, the time slice that is positioned at time series first end be set to first reference time fragment.
Wherein, time series two ends can be defined as first end and the second end, first end can be time series foremost, can be also the rearmost end of time series.
Step B, judge first reference time fragment whether be the time slice of time series the second end, if so, process ends, otherwise, first reference time fragment the distolateral adjacent time slice of close time series second be set to second reference time fragment.
Wherein, if definition time series first end be time series foremost, the second end is the rearmost end of time series, if first end is the rearmost end of time series, the second end be time series foremost.First reference time fragment the distolateral adjacent time slice of close time series second, be the time slice nearer with adjacent time slice middle distance time series the second end of the first fragment reference time.
Step C, judges whether the absolute value of the slope differences of the first line segment corresponding to fragment reference time and line segment corresponding to the second fragment reference time is less than default first threshold, if so, performs step D, otherwise, execution step E.
Wherein, the value of first threshold can take with requirements such as processing speeds and arrange according to degree of accuracy, storage space.The slope of two line segments approaches, and illustrates that the straight line at these two line segment places is seemingly closer, if merged, less on the impact of data accuracy, and can reduce storage space, takies.
Step D, judge second reference time fragment whether be the time slice of time series the second end, if, by first reference time fragment to the second fragment reference time a plurality of time slices that are linked in sequence merge processing, and process ends, otherwise, second reference time fragment the distolateral adjacent time slice of close time series second be set to second reference time fragment, and go to execution step C.
Concrete, by first reference time fragment to the second fragment reference time a plurality of time slices that are linked in sequence merge in the process of processing, if first reference time fragment and second reference time fragment be adjacent time slice, can be to first reference time fragment and second reference time fragment merge processing, if first reference time fragment and second reference time fragment be not adjacent time slice, can be to first reference time fragment and second reference time fragment and the time slice between them merge processing.
Step e, judge first reference time fragment be whether second reference time fragment adjacent time slice, if, second reference time fragment be set to first reference time fragment, and go to execution step B, otherwise, by comprise first reference time fragment and the second reference time fragment between time slice and first reference time fragment a plurality of time slices that are linked in sequence merge processing, second reference time fragment be set to first reference time fragment, and go to execution step B.
In above-mentioned flow process, from one end of time series, start to the other end, the line segment by the line segment in each time slice and first time slice of order carries out slope ratio, until the slope differences of current time fragment and first time slice is while reaching first threshold, first time slice is merged to processing up to the previous time slice of current time fragment, then, current time fragment is repeated to process above as first time slice, until the time slice of the time series other end.
Below in conjunction with a concrete example, above-mentioned flow process is described: supposing that time series is divided into the time slice that 5 durations are unit duration, is respectively that time slice 1 is to time slice 5 from front to back.First, the line segment in the line segment in time slice 2 and time slice 1 can be carried out to slope ratio, suppose that slope differences is less than first threshold; Then, continue the line segment in the line segment in time slice 3 and time slice 1 to carry out slope ratio, suppose that slope differences is still less than first threshold; Again, continue the line segment in the line segment in time slice 4 and time slice 1 to carry out slope ratio, suppose that at this moment slope differences is greater than first threshold, at this moment, time slice 1, time slice 2 and time slice 3 are merged to processing; Again, from time slice 4, the line segment in the line segment in time slice 5 and time slice 4 is carried out to slope ratio, suppose that slope differences is less than first threshold, at this moment, time slice 4 and time slice 5 are merged to processing process ends.
Processing mode two, its concrete implementation can comprise the steps:
Step H, is in each time slice of unit duration at duration, the time slice that is positioned at time series first end be set to the 3rd reference time fragment.
Step I, judge the 3rd reference time fragment whether be the time slice of time series the second end, if so, process ends, otherwise, execution step J.
Step J, judge the 3rd reference time fragment line segment corresponding to the distolateral adjacent time slice of close time series second whether be less than default first threshold with the absolute value of the slope differences of line segment corresponding to the 3rd fragment reference time, if so, perform step K, otherwise, execution step L.
Wherein, the value of first threshold can take with requirements such as processing speeds and arrange according to degree of accuracy, storage space.
Step K, by comprise the 3rd reference time fragment the distolateral adjacent time slice of close time series second and the 3rd reference time fragment a plurality of time slices that are linked in sequence merge processing, the time slice that merging obtains be set to the 3rd reference time fragment, and go to execution step I.
Step L, the 3rd reference time fragment the distolateral adjacent time slice of close time series second be set to the 3rd reference time fragment, and go to execution step I.
In above-mentioned steps, from one end of time series, start to the other end, compare one by one the slope of time slice middle conductor, if the slope differences of the line segment in adjacent time slice is less than first threshold, time slice is merged to processing, and the time slice after use merging and the time slice of rear adjacent carry out slope ratio, by that analogy.
Below in conjunction with a concrete example, above-mentioned flow process is described: supposing that time series is divided into the time slice that 5 durations are unit duration, is respectively that time slice 1 is to time slice 5 from front to back.First, the line segment in the line segment in time slice 2 and time slice 1 can be carried out to slope ratio, suppose that slope differences is less than first threshold, time slice 1 and time slice 2 be merged process and obtain time slice 2 '; Then, the line segment in the line segment in time slice 3 and time slice 2 ' is carried out to slope ratio, suppose that slope differences is still less than first threshold, time slice 2 ' and time slice 3 are merged to processing and obtain time slice 3 '; Again, line segment in line segment in time slice 4 and time slice 3 ' is carried out to slope ratio, suppose that at this moment slope differences is greater than first threshold, from time slice 4, start to compare, line segment in line segment in time slice 5 and time slice 4 is carried out to slope ratio, suppose that slope differences is less than first threshold, at this moment, time slice 4 and time slice 5 are merged to processing process ends.
Method two, can comprise following treatment step:
Step 1, in time range corresponding to the duration of the time series obtaining, determine time point, these Time intervals are the integral multiple (these time points can be called unit interval point) of default unit duration from the duration of the start time point of time series or termination time point, and the duration of time series is the integral multiple of unit duration.
Because the duration of time series is the integral multiple of unit duration, so be the time point of integral multiple and the integral multiple of time series termination time point duration Ye Shi unit duration apart of unit duration apart from the duration of time series start time point.
Step 2, the crest comprising according to the waveform of time series and the time point of trough, in the above-mentioned time point of determining access time fragment boundary time point.
Concrete, time point for each crest, can be in above-mentioned each time point of determining, the distance of choosing the time point of one or more and crest is less than the time point of certain default duration, time point for each trough, can be in above-mentioned each time point of determining, the distance of choosing the time point of one or more and trough is less than the time point of certain default duration, the boundary time point using the time point of choosing as time slice.Like this access time fragment boundary time point carry out PLA processing, can guarantee to carry out time slice division according to data variation trend as far as possible, when saving storage space, assurance data accuracy that can be to a certain degree.
Preferably, can be in the above-mentioned time point of determining, choose the time point that is less than described unit duration with the distance of the time point of each crest, and be less than the time point of described unit duration the boundary time point using the time point of choosing as time slice with the distance of the time point of each trough.Concrete, for the time point of each crest and trough, can choose at least one and its unit interval point apart from the unit's of being less than duration, as the boundary time point of time slice.Time point for a crest or trough, if overlapped with certain unit interval point, in constituent parts time point, put and only have one from the unit interval of the unit's of being less than duration with the Time interval of this crest or trough, the unit interval point for this reason overlapping, can select this unit interval point as the boundary time point of time slice.Time point for a crest or trough, if be positioned in the middle of two unit interval points, in constituent parts time point, put and have two from the unit interval of the unit's of being less than duration with the Time interval of this crest or trough, be this two unit interval points, can select any boundary time point as time slice wherein, also these two unit interval points all can be elected to be to the boundary time point of time slice.
Preferably, can be in the above-mentioned time point of determining, choose the time point with the distance minimum of the time point of each crest, and with the time point of the distance minimum of the time point of each trough, and the boundary time point using the time point of choosing as time slice.Concrete, for the time point of each crest and trough, can choose and its unit interval point apart from minimum, as the boundary time point of time slice.Like this access time fragment boundary time point carry out PLA processing, can be more effective the data accuracy of the time series of assurance PLA after processing.
Step 3, the boundary time point according to the time slice of choosing, carries out PLA processing to time series.
This method, processing procedure is comparatively simple, can effectively improve the treatment effeciency of step 102.
The embodiment of the present invention also provides a kind of method that time series is retrieved.In the method, can be pre-stored through the time series of dimension-reduction treatment, the mode of this dimension-reduction treatment can be any dimension-reduction treatment mode of being scheduled to, as conventional P LA processes, fixed length PLA processes etc.Preferably, time series is carried out to the time series that the method for dimension-reduction treatment is processed in can pre-stored employing above-described embodiment, concrete processing procedure can, with reference to the detailed description to step 102 above, be not repeated at this.As shown in Figure 2, the treatment scheme of the method can comprise following step:
Step 201, receives the inquiry request of carrying Goal time order sequence.Wherein, the duration of the duration of Goal time order sequence and the time series of storage is identical.
Step 202, the same way that employing is carried out dimension-reduction treatment to the time series of storage is carried out dimension-reduction treatment to Goal time order sequence.If the method that time series is carried out to dimension-reduction treatment in employing above-described embodiment is carried out dimension-reduction treatment to Goal time order sequence, concrete processing procedure can, with reference to the detailed description to step 102 above, be not repeated at this.
Step 203, in the time series of storage, the time series that the Goal time order sequence after inquiry and processing matches.
Wherein, the Goal time order sequence after processing and each time series of storage have recorded the data of a series of line segments, so can be expressed as a broken line.Goal time order sequence after processing and the time series of storage match, and are also that the degree of approximation of their data reaches certain requirement, and the degree of approximation can be understood as between the broken line of their correspondences has reached certain requirement.Concrete, the decision procedure of coupling can be Goal time order sequence after judgement is processed with the time series of storage between the absolute value of distance whether be less than certain default distance threshold, if so, judge coupling, otherwise judgement is not mated.
Between two time seriess, the computing method of distance have a variety of, preferably, can adopt following method to calculate distance: the area of the figure that the Goal time order sequence broken line corresponding with the time series of storage after computing forms within the scope of their durations, here, distance between definition time series exists positive and negative, the area of figure also exists positive and negative so, accordingly.Concrete, each enclosure portion for figure, if the mean value of definition the first time series is greater than the mean value of the second time series, the area of this part is being for just, otherwise area is for negative, also can be conversely, each enclosure portion for figure, if the mean value of definition the first time series is greater than the mean value of the second time series, the area of this part is for negative, otherwise area is for just.For an enclosure portion of figure, in the time range of this enclosure portion, two broken lines that time series is corresponding, which broken line illustrates that the mean value of the time series which broken line is corresponding is larger in the above.Calculate after the area of each enclosure portion of figure, these area numerical value have just to be had negatively, and the area numerical value of these enclosure portion is added, and can obtain the area of whole figure, and the area of this figure is distance between these two time seriess.
In embodiments of the present invention, can also record first distance corresponding to each time series of storage, wherein, the first distance be time series after processing with default benchmark time series between distance, first apart from being defined as reference distance again.Based on this, the implementation of above-mentioned steps 203 can be: obtain second distance, second distance is Goal time order sequence and the distance between benchmark time series (second distance can be defined as target range again) after processing; In the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance be less than default Second Threshold (range difference that obtains reference distance and target range is less than the time series of default Second Threshold), as with processing after the time series that matches of Goal time order sequence.
Wherein, the duration of benchmark time series is identical with the duration of the time series of storage, time series data in benchmark time series can arrange arbitrarily, preferably, for the ease of calculating distance, improve treatment effeciency, benchmark time series can be set and only include a time slice, and the slope of line segment corresponding to this time slice is 0.Reference distance calculates after can the time series after processing storing in database, without calculating in the process in inquiry again.
The range difference of reference distance and target range, is the time series of corresponding storage and the distance between the Goal time order sequence after processing.In above-mentioned query script, only need the simple range difference that calculates each reference distance and target range, just can determine time series after corresponding processing and process after Goal time order sequence between distance, after comparing with default Second Threshold, just can determine qualified time series, with process after the time series that matches of Goal time order sequence.With respect to prior art the Goal time order sequence after in real time computing and the distance between each time series of storage in query script, the method that the embodiment of the present invention provides can effectively promote the treatment effeciency of query script.
Preferably, can also set up R-tree to the time series of storage, and record each the MBR(Minimum Bounding Rectangle in R-tree, minimum boundary rectangle) corresponding minimum frontier distance, wherein, minimum frontier distance is the minimum value of the first distance (reference distance) of each time series in MBR.
R-tree is a kind of data description mode based on tree construction, is mainly used to improve by index data the search efficiency of these data.The core concept of R-tree is that mutually similar data object is grouped in together, and every group of minimum boundary rectangle for data object (MBR) described to this group data object on the tree node of higher level.Because all similar data objects are included in a MBR, in Query Database, match with target data object data object time, only need to be from top, in every layer, search the MBR that can comprise target data object, until then the bottom searches the data object matching with target data object in each data object in the MBR of the bottom.Above-mentioned data object can be time series, and R-tree can be described for the time series in data storehouse.Reference distance and minimum frontier distance can carry out record in R-tree.
When the time series after having new processing deposits database in, join in the scope of certain MBR, can calculate the reference distance of this time series, and can determine whether and need to upgrade the minimum frontier distance of this MBR according to this reference distance, if this reference distance is less than the minimum frontier distance of this MBR, the minimum frontier distance of this MBR is updated to the numerical value of this reference distance.If this MBR is full, can be divided into two MBR, this time series is inserted to one of them MBR, then record the minimum frontier distance of two MBR.
Minimum frontier distance based on R-tree and each MBR, the time series that the above-mentioned range difference that obtains the first corresponding distance and second distance in the time series of storage is less than default Second Threshold as with process after the processing procedure of the time series that matches of Goal time order sequence, can be: according to corresponding minimum frontier distance and the first distance corresponding to time series of MBR in the R-tree of record, in the time series of storage, obtain the time series that the first corresponding distance and the range difference of second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.Concrete processing procedure can comprise: the range difference that obtains minimum frontier distance and second distance in R-tree is less than the MBR of default Second Threshold; In each time series in the MBR obtaining, obtain the time series that the first corresponding distance and the range difference of second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
In implementation process, from R-tree top, in every layer outside the bottom of R-tree, the range difference of inquiring about minimum frontier distance and target range is less than the MBR of default Second Threshold, and in the MBR inquiring, carries out the inquiry of lower one deck; In the bottom of R-tree, the range difference that obtains reference distance and target range is less than the time series of default Second Threshold, as with processing after the time series that matches of Goal time order sequence.
In above-mentioned query script, only needs are simple calculates the minimum frontier distance of MBR and the difference of target range, and the difference of calculating corresponding reference distance and target range, just can carry out the query script of R-tree.With respect to prior art, in R-tree query process, calculate in real time the mode of distance, the method that the embodiment of the present invention provides can effectively improve the efficiency of R-tree query.
The process of the R-tree query in said method can realize by program below:
Wherein, Q is the Goal time order sequence after PLA processes, and ε is default Second Threshold, and b is benchmark time series.The Output rusults of this program can be result={TS1, TS2, and TS3 ..., TSn}.
In the embodiment of the present invention, adopt the duration of time slice fixing and be the PLA processing mode of the integral multiple of default unit duration, time series is carried out to dimension-reduction treatment, like this, with respect to fixed length PLA, can substitute a plurality of time slices in fixed length PLA with a time slice, thereby, storage space shared while storing time series can be reduced.
Technical conceive based on identical, the embodiment of the present invention also provides a kind of device that time series is carried out to dimension-reduction treatment, and as shown in Figure 3, described device comprises:
Acquisition module 310, for obtaining pending time series;
Processing module 320, for described time series being carried out to piecewise linear approximation PLA processing, the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration.
Preferably, described processing module 320, specifically for:
The duration that described time series is carried out to time slice is that the PLA of default unit duration processes, and the duration of described time series is the integral multiple of described unit duration;
At duration, be in each time slice of described unit duration, if meet default approximate condition between line segment corresponding to a plurality of time slices that are linked in sequence, the described a plurality of time slices that are linked in sequence merged to processing.
Preferably, described processing module 320, specifically for:
At duration, be in each time slice of described unit duration, if the absolute value of line segment corresponding to a plurality of time slices that are linked in sequence slope differences is each other less than default first threshold, the described a plurality of time slices that are linked in sequence merged to processing.
Preferably, described processing module 320, specifically for:
In time range corresponding to the duration of described time series, determine time point, described Time interval is the integral multiple of default unit duration from the duration of the start time point of described time series or termination time point, and the duration of described time series is the integral multiple of described unit duration;
The crest comprising according to the waveform of described time series and the time point of trough, in the time point of determining access time fragment boundary time point;
Boundary time point according to the time slice of choosing, carries out PLA processing to described time series.
Preferably, described processing module 320, specifically for:
In the described time point of determining, choose the time point that is less than described unit duration with the distance of the time point of each crest, and be less than the time point of described unit duration the boundary time point using the time point of choosing as time slice with the distance of the time point of each trough.
Preferably, described processing module 320, specifically for:
In the described time point of determining, choose the time point with the distance minimum of the time point of each crest, and with the time point of the distance minimum of the time point of each trough, and the boundary time point using the time point of choosing as time slice.
Preferably, also comprise: memory module, for the time series after processing is stored.
Technical conceive based on identical, the embodiment of the present invention also provides a kind of device that time series is retrieved, and as shown in Figure 4, described device comprises:
Memory module 410, for the pre-stored time series through dimension-reduction treatment, the mode of this dimension-reduction treatment can be any dimension-reduction treatment mode of being scheduled to, as conventional P LA processing, fixed length PLA processing etc., preferably, this memory module 410 is for the pre-stored employing time series that time series is carried out to the device processing of dimension-reduction treatment described above;
Receiver module 420, for receiving the inquiry request of carrying Goal time order sequence;
Processing module 430, for adopting the same way that the time series of storage is carried out to dimension-reduction treatment to carry out dimension-reduction treatment to described Goal time order sequence;
Enquiry module 440, for the time series in storage, the time series that the Goal time order sequence after inquiry and processing matches.
Preferably, described memory module 410, also for:
First distance corresponding to each time series of record storage, wherein, the distance between the time series that described the first distance is described storage and default benchmark time series;
Described enquiry module 440, for:
Obtain second distance, described second distance is Goal time order sequence after described processing and the distance between described benchmark time series;
In the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
Preferably, described memory module 410, also for:
The time series of storage is set up to R-tree; Record minimum frontier distance corresponding to each minimum boundary rectangle MBR in described R-tree, wherein, described minimum frontier distance is the minimum value of the first distance of each time series in described MBR;
Described enquiry module 440, for:
According to corresponding minimum frontier distance and the first distance corresponding to time series of MBR in the described R-tree of record, in the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
Preferably, described enquiry module 440, for:
The range difference that obtains minimum frontier distance and described second distance in described R-tree is less than the MBR of default Second Threshold;
In each time series in the MBR obtaining, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
In the embodiment of the present invention, adopt the duration of time slice fixing and be the PLA processing mode of the integral multiple of default unit duration, time series is carried out to dimension-reduction treatment, like this, with respect to fixed length PLA, can substitute a plurality of time slices in fixed length PLA with a time slice, thereby, storage space shared while storing time series can be reduced.
It should be noted that: the device that time series is carried out to dimension-reduction treatment that above-described embodiment provides is when carrying out dimension-reduction treatment to time series, only the division with above-mentioned each functional module is illustrated, in practical application, can above-mentioned functions be distributed and by different functional modules, completed as required, the inner structure that is about to device is divided into different functional modules, to complete all or part of function described above.In addition, the device that time series is carried out to dimension-reduction treatment that above-described embodiment provides belongs to same design with the embodiment of the method for time series being carried out to dimension-reduction treatment, and its specific implementation process refers to embodiment of the method, repeats no more here.
The invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.
One of ordinary skill in the art will appreciate that all or part of step that realizes above-described embodiment can complete by hardware, also can come the hardware that instruction is relevant to complete by program, described program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium of mentioning can be ROM (read-only memory), disk or CD etc.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.
Claims (22)
1. time series is carried out to a method for dimension-reduction treatment, it is characterized in that, described method comprises:
Obtain pending time series;
Described time series is carried out to piecewise linear approximation PLA processing, and the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration.
2. method according to claim 1, is characterized in that, described described time series is carried out to PLA processing, and the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration, comprising:
The duration that described time series is carried out to time slice is that the PLA of default unit duration processes, and the duration of described time series is the integral multiple of described unit duration;
At duration, be in each time slice of described unit duration, if meet default approximate condition between line segment corresponding to a plurality of time slices that are linked in sequence, the described a plurality of time slices that are linked in sequence merged to processing.
3. method according to claim 2, it is characterized in that, described is in each time slice of described unit duration at duration, if meet default approximate condition between line segment corresponding to a plurality of time slices that are linked in sequence, the described a plurality of time slices that are linked in sequence are merged to processing, comprising:
At duration, be in each time slice of described unit duration, if the absolute value of line segment corresponding to a plurality of time slices that are linked in sequence slope differences is each other less than default first threshold, the described a plurality of time slices that are linked in sequence merged to processing.
4. method according to claim 1, is characterized in that, described described time series is carried out to PLA processing, and the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration, comprising:
In time range corresponding to the duration of described time series, determine time point, described Time interval is the integral multiple of default unit duration from the duration of the start time point of described time series or termination time point, and the duration of described time series is the integral multiple of described unit duration;
The crest comprising according to the waveform of described time series and the time point of trough, in the time point of determining access time fragment boundary time point;
Boundary time point according to the time slice of choosing, carries out PLA processing to described time series.
5. method according to claim 4, is characterized in that, the time point of the described crest comprising according to the waveform of described time series and trough, in the time point of determining access time fragment boundary time point, comprising:
In the described time point of determining, choose the time point that is less than described unit duration with the distance of the time point of each crest, and be less than the time point of described unit duration the boundary time point using the time point of choosing as time slice with the distance of the time point of each trough.
6. method according to claim 4, is characterized in that, the time point of the described crest comprising according to the waveform of described time series and trough, in the time point of determining access time fragment boundary time point, comprising:
In the described time point of determining, choose the time point with the distance minimum of the time point of each crest, and with the time point of the distance minimum of the time point of each trough, and the boundary time point using the time point of choosing as time slice.
7. according to the method described in claim 1-6 any one, it is characterized in that, described described time series is carried out to piecewise linear approximation PLA processing, the duration of the time slice that described PLA processes duration fixing and described time slice is after the integral multiple of default unit duration, also comprises:
Time series after processing is stored.
8. a method of time series being retrieved, is characterized in that, pre-stored have to adopt as described in claim 1-7, time series is carried out to the time series that the method for dimension-reduction treatment is processed, described method comprises:
The inquiry request of Goal time order sequence is carried in reception;
The same way that employing is carried out dimension-reduction treatment to the time series of storage is carried out dimension-reduction treatment to described Goal time order sequence;
In the time series of storage, the time series that the Goal time order sequence after inquiry and processing matches.
9. method according to claim 8, is characterized in that, also comprises:
First distance corresponding to each time series of record storage, wherein, the distance between the time series that described the first distance is described storage and default benchmark time series;
The time series described in the time series of storage, the Goal time order sequence after inquiry and processing matches, comprising:
Obtain second distance, described second distance is Goal time order sequence after described processing and the distance between described benchmark time series;
In the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
10. method according to claim 9, is characterized in that, also comprises:
The time series of storage is set up to R-tree; Record minimum frontier distance corresponding to each minimum boundary rectangle MBR in described R-tree, wherein, described minimum frontier distance is the minimum value of the first distance of each time series in described MBR;
Described in the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence, comprising:
According to corresponding minimum frontier distance and the first distance corresponding to time series of MBR in the described R-tree of record, in the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
11. methods according to claim 10, it is characterized in that, according to corresponding minimum frontier distance and the first distance corresponding to time series of MBR in the described R-tree of record, in the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence, comprising:
The range difference that obtains minimum frontier distance and described second distance in described R-tree is less than the MBR of default Second Threshold;
In each time series in the MBR obtaining, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
12. 1 kinds of devices that time series carried out to dimension-reduction treatment, is characterized in that, described device comprises:
Acquisition module, for obtaining pending time series;
Processing module, for described time series being carried out to piecewise linear approximation PLA processing, the duration of the time slice that described PLA processes duration fixing and described time slice is the integral multiple of default unit duration.
13. devices according to claim 12, is characterized in that, described processing module, specifically for:
The duration that described time series is carried out to time slice is that the PLA of default unit duration processes, and the duration of described time series is the integral multiple of described unit duration;
At duration, be in each time slice of described unit duration, if meet default approximate condition between line segment corresponding to a plurality of time slices that are linked in sequence, the described a plurality of time slices that are linked in sequence merged to processing.
14. devices according to claim 13, is characterized in that, described processing module, specifically for:
At duration, be in each time slice of described unit duration, if the absolute value of line segment corresponding to a plurality of time slices that are linked in sequence slope differences is each other less than default first threshold, the described a plurality of time slices that are linked in sequence merged to processing.
15. devices according to claim 12, is characterized in that, described processing module, specifically for:
In time range corresponding to the duration of described time series, determine time point, described Time interval is the integral multiple of default unit duration from the duration of the start time point of described time series or termination time point, and the duration of described time series is the integral multiple of described unit duration;
The crest comprising according to the waveform of described time series and the time point of trough, in the time point of determining access time fragment boundary time point;
Boundary time point according to the time slice of choosing, carries out PLA processing to described time series.
16. devices according to claim 15, is characterized in that, described processing module, specifically for:
In the described time point of determining, choose the time point that is less than described unit duration with the distance of the time point of each crest, and be less than the time point of described unit duration the boundary time point using the time point of choosing as time slice with the distance of the time point of each trough.
17. devices according to claim 15, is characterized in that, described processing module, specifically for:
In the described time point of determining, choose the time point with the distance minimum of the time point of each crest, and with the time point of the distance minimum of the time point of each trough, and the boundary time point using the time point of choosing as time slice.
18. according to the device described in claim 12-17 any one, it is characterized in that, also comprises:
Memory module, for storing the time series after processing.
19. 1 kinds of devices that time series is retrieved, is characterized in that, described device comprises:
Memory module, the time series that time series is carried out to the device processing of dimension-reduction treatment for pre-stored employing as described in claim 12-18;
Receiver module, for receiving the inquiry request of carrying Goal time order sequence;
Processing module, for adopting the same way that the time series of storage is carried out to dimension-reduction treatment to carry out dimension-reduction treatment to described Goal time order sequence;
Enquiry module, for the time series in storage, the time series that the Goal time order sequence after inquiry and processing matches.
20. devices according to claim 19, is characterized in that, described memory module, also for:
First distance corresponding to each time series of record storage, wherein, the distance between the time series that described the first distance is described storage and default benchmark time series;
Described enquiry module, for:
Obtain second distance, described second distance is Goal time order sequence after described processing and the distance between described benchmark time series;
In the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
21. devices according to claim 20, is characterized in that, described memory module, also for:
The time series of storage is set up to R-tree; Record minimum frontier distance corresponding to each minimum boundary rectangle MBR in described R-tree, wherein, described minimum frontier distance is the minimum value of the first distance of each time series in described MBR;
Described enquiry module, for:
According to corresponding minimum frontier distance and the first distance corresponding to time series of MBR in the described R-tree of record, in the time series of storage, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
22. devices according to claim 21, is characterized in that, described enquiry module, for:
The range difference that obtains minimum frontier distance and described second distance in described R-tree is less than the MBR of default Second Threshold;
In each time series in the MBR obtaining, obtain the time series that the first corresponding distance and the range difference of described second distance are less than default Second Threshold, as with process after the time series that matches of Goal time order sequence.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310105366.0A CN104077309B (en) | 2013-03-28 | 2013-03-28 | A kind of method and apparatus that dimension-reduction treatment is carried out to time series |
| JP2014043463A JP2014194762A (en) | 2013-03-28 | 2014-03-06 | Method and device for processing time sequence based on dimensionality reduction |
| US14/203,852 US9396247B2 (en) | 2013-03-28 | 2014-03-11 | Method and device for processing a time sequence based on dimensionality reduction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310105366.0A CN104077309B (en) | 2013-03-28 | 2013-03-28 | A kind of method and apparatus that dimension-reduction treatment is carried out to time series |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104077309A true CN104077309A (en) | 2014-10-01 |
| CN104077309B CN104077309B (en) | 2018-05-08 |
Family
ID=51598569
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310105366.0A Active CN104077309B (en) | 2013-03-28 | 2013-03-28 | A kind of method and apparatus that dimension-reduction treatment is carried out to time series |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US9396247B2 (en) |
| JP (1) | JP2014194762A (en) |
| CN (1) | CN104077309B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021579A (en) * | 2016-06-01 | 2016-10-12 | 南京国电南自美卓控制系统有限公司 | Compression method of historical database |
| CN107918914A (en) * | 2017-11-02 | 2018-04-17 | 浙江富帝科技有限公司 | A kind of time series data segmentation method in banking software |
| CN116269258A (en) * | 2023-02-22 | 2023-06-23 | 昆山市第一人民医院 | Pillow type sleep detection method and device and computer equipment |
| CN116383273A (en) * | 2023-04-04 | 2023-07-04 | 济南永信新材料科技有限公司 | Time sequence dimension reduction representation method and system |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6345346B2 (en) * | 2015-05-28 | 2018-06-20 | 三菱電機株式会社 | Analysis apparatus, analysis method, and analysis program |
| CN106600449B (en) * | 2015-10-16 | 2021-03-16 | 国家电网公司 | Automatic power trend identification method |
| CN107391148B (en) * | 2017-07-28 | 2020-09-15 | 广州视源电子科技股份有限公司 | View element saving method and device, electronic equipment and computer storage medium |
| US10891335B2 (en) | 2018-01-03 | 2021-01-12 | International Business Machines Corporation | Enhanced exploration of dimensionally reduced data |
| CN109376953B (en) * | 2018-11-21 | 2022-09-23 | 国网冀北电力有限公司 | Medium-and-long-term electric quantity prediction method and system |
| CN110620586B (en) * | 2019-09-17 | 2021-07-13 | 河北省科学院应用数学研究所 | Data compression method and device and terminal equipment |
| CN111104966B (en) * | 2019-11-30 | 2023-04-18 | 西安外事学院 | Two-dimensional closed graph dimension reduction method based on time sequence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080071843A1 (en) * | 2006-09-14 | 2008-03-20 | Spyridon Papadimitriou | Systems and methods for indexing and visualization of high-dimensional data via dimension reorderings |
| CN101763417A (en) * | 2009-12-30 | 2010-06-30 | 北京世纪高通科技有限公司 | Data query method and device |
| US20120162293A1 (en) * | 2006-04-27 | 2012-06-28 | Edwin Lyle Hudson | Gray scale drive sequences for pulse width modulated displays |
| CN102682573A (en) * | 2012-03-31 | 2012-09-19 | 上海海洋大学 | Time sequence analysis-based storm surge disaster early warning system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3562572B2 (en) * | 2000-05-02 | 2004-09-08 | インターナショナル・ビジネス・マシーンズ・コーポレーション | Detect and track new items and new classes in database documents |
| JP5095109B2 (en) * | 2006-03-09 | 2012-12-12 | 三菱電機株式会社 | Tracking device |
| US8219574B2 (en) * | 2009-06-22 | 2012-07-10 | Microsoft Corporation | Querying compressed time-series signals |
-
2013
- 2013-03-28 CN CN201310105366.0A patent/CN104077309B/en active Active
-
2014
- 2014-03-06 JP JP2014043463A patent/JP2014194762A/en active Pending
- 2014-03-11 US US14/203,852 patent/US9396247B2/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120162293A1 (en) * | 2006-04-27 | 2012-06-28 | Edwin Lyle Hudson | Gray scale drive sequences for pulse width modulated displays |
| US20080071843A1 (en) * | 2006-09-14 | 2008-03-20 | Spyridon Papadimitriou | Systems and methods for indexing and visualization of high-dimensional data via dimension reorderings |
| CN101763417A (en) * | 2009-12-30 | 2010-06-30 | 北京世纪高通科技有限公司 | Data query method and device |
| CN102682573A (en) * | 2012-03-31 | 2012-09-19 | 上海海洋大学 | Time sequence analysis-based storm surge disaster early warning system |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021579A (en) * | 2016-06-01 | 2016-10-12 | 南京国电南自美卓控制系统有限公司 | Compression method of historical database |
| CN106021579B (en) * | 2016-06-01 | 2019-11-08 | 南京国电南自维美德自动化有限公司 | A kind of compression method of historical data base |
| CN107918914A (en) * | 2017-11-02 | 2018-04-17 | 浙江富帝科技有限公司 | A kind of time series data segmentation method in banking software |
| CN116269258A (en) * | 2023-02-22 | 2023-06-23 | 昆山市第一人民医院 | Pillow type sleep detection method and device and computer equipment |
| CN116269258B (en) * | 2023-02-22 | 2023-11-21 | 昆山市第一人民医院 | Pillow type sleep detection method and device and computer equipment |
| CN116383273A (en) * | 2023-04-04 | 2023-07-04 | 济南永信新材料科技有限公司 | Time sequence dimension reduction representation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| US20140297606A1 (en) | 2014-10-02 |
| JP2014194762A (en) | 2014-10-09 |
| US9396247B2 (en) | 2016-07-19 |
| CN104077309B (en) | 2018-05-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104077309A (en) | Method and device for carrying out dimension reduction processing on time-sequential sequence | |
| CN109189991B (en) | Duplicate video identification method, device, terminal and computer readable storage medium | |
| CN105389349B (en) | Dictionary update method and device | |
| US9020271B2 (en) | Adaptive hierarchical clustering algorithm | |
| US9361343B2 (en) | Method for parallel mining of temporal relations in large event file | |
| CN103248705B (en) | Server, client and method for processing video frequency | |
| CN106530132A (en) | Power load clustering method and device | |
| CN109326325A (en) | A kind of method, system and associated component that gene order compares | |
| CN110598065A (en) | Data mining method and device and computer readable storage medium | |
| CN106294390A (en) | A kind of data mining analysis method and system | |
| CN103345616A (en) | Fingerprint storage comparison system based on behavioral analysis | |
| CN107623639A (en) | Data flow distribution similarity join method based on EMD distances | |
| CN104902292A (en) | Television report-based public opinion analysis method and system | |
| CN105488134A (en) | Big data processing method and big data processing device | |
| US8392393B2 (en) | Graph searching | |
| CN101477688B (en) | Method and apparatus for extracting edge of target | |
| CN102436535B (en) | Identification method and system for creative inflection point in computer aided design process | |
| CN110032585B (en) | Time sequence double-layer symbolization method and device | |
| CN104778202B (en) | The analysis method and system of event evolutionary process based on keyword | |
| CN112445833A (en) | Data paging query method, device and system for distributed database | |
| CN112148942A (en) | Business index data classification method and device based on data clustering | |
| CN112347813A (en) | Baseline detection method, equipment and storage medium for high signal-to-noise ratio time series | |
| CN104516916A (en) | Method and device for analyzing network report incidence relation | |
| US20170193081A1 (en) | Intersection algorithm-based search method and system, and storage medium | |
| CN114298992A (en) | Video frame duplication removing method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |